Aleksandr Razumov
Aleksandr Razumov
Improve performance of LogQL, TraceQL, PromQL common queries to
Implement Kafka receiver for logs, metrics and traces. Format is protobuf, same as kafka exporter.
Memory usage spikes to ~gigabyte+, CPU usage spikes to all available cores during viewing last 12 hours on node exporter. Available on "demo" dev local stand. Metrics: https://storage.yandexcloud.net/faster-public/dump.metrics.full.native.zstd Grafana link:...
``` sum(rate({"chstorage.inserts", "chstorage.signal"="$signal"}[$__rate_interval])) ```
``` --- FAIL: TestInstantAggregation (0.02s) --- FAIL: TestInstantAggregation/Test12 (0.00s) query_test.go:178: Error Trace: /home/ubuntu/oteldb/internal/logql/logqlengine/logqlmetric/query_test.go:178 Error: Not equal: expected: "2.9899999999999998" actual : "2.99" Diff: --- Expected +++ Actual @@ -1 +1 @@...
Refactor helm chart to use structured values for ClickHouse DSN, etc instead of using environment variables.
Yet another approach for materialized fields. 1. Materialize subset of `resource`. E.g. `resource['application']` into `resource_application` 2. Alter queries to use materialized field when querying on this resource or attribute 3....
Example resource: ```json {"k8s.container.name":"oteldb","k8s.container.restart_count":"7","k8s.deployment.name":"chotel","k8s.namespace.name":"faster","k8s.node.name":"dev-idp-mgmt-md-0-r7r4j-f4wfj-7n6r6","k8s.pod.name":"chotel-57449689f-89chp","k8s.pod.start_time":"2025-10-16T20:51:30Z","k8s.pod.uid":"6fbd684d-e987-46ca-96de-826e6f590cc9","service.instance.id":"6fbd684d-e987-46ca-96de-826e6f590cc9"} ``` Probably we can automatically pick `k8s.deployment.name` as service name, loki is doing something similar.